Financial constraint and cognitive load: one finding, one non-replication

In 2013, Science published a paper titled “Poverty Impedes Cognitive Function,” by Anandi Mani, Sendhil Mullainathan, Eldar Shafir and Jiaying Zhao. It has since become one of the most cited empirical claims in behavioural economics, and it is frequently summarised in ways that overstate what it established.

The paper reported two sets of results. The first came from experiments run with shoppers at a New Jersey mall. Participants were asked to consider a hypothetical car repair — assigned at random to be either inexpensive or expensive — and then completed standard cognitive measures, including Raven's Progressive Matrices and a cognitive control task. Among lower-income participants, performance dropped after the expensive scenario. Among higher-income participants, it did not. The second set came from a field study of several hundred sugarcane farmers in Tamil Nadu, whose income arrives in a single annual lump at harvest. The same farmers were tested before harvest, when cash was short, and after harvest, when it was not. Post-harvest scores were higher. The authors compared the size of the pre-harvest gap to the cognitive cost of a full night without sleep.

The proposed mechanism is attentional rather than dispositional. Under this account, a binding financial constraint is not simply an external circumstance; it occupies working memory, in the way an unresolved problem occupies attention during a conversation. The claim is about cognitive load, not about character or intelligence.

What the replication attempt found

In 2016, the American Economic Review published “Poverty and Economic Decision-Making: Evidence from Households on Payday,” by Leandro Carvalho, Stephan Meier and Stephanie Wang. The design used naturally occurring variation in liquidity: low-income households were surveyed either shortly before or shortly after payday, with timing assigned at random. On cognitive measures, including Raven's matrices and a numerical Stroop task, the study found no difference between the two groups. It did report differences in intertemporal choice — respondents surveyed before payday behaved more impatiently in some decisions.

The two studies are not straightforwardly comparable, and the authors of the later paper say so. A payday cycle is short, anticipated and modest in amplitude. An annual harvest cycle for a farming household is none of those things. A reasonable reading is that the bandwidth effect, if it exists, requires a constraint of some depth and duration, and does not appear in the ordinary two-week rhythm of a paycheque.

The statistical objections

Separately, Jelte Wicherts and Annemarie Scholten raised methodological concerns about the farmer study shortly after publication, focused on the repeated-measures design. Participants sat the same class of test twice, months apart, which introduces practice effects that run in the same direction as the reported result. The sample, while substantial for field work of this kind, is small relative to the precision often implied when the finding is restated in summary form.

The comparison to lost sleep, expressed in IQ-point equivalents, has also drawn criticism as a communication device. Converting a task-specific score difference into a general intelligence metric implies a stability of measurement that the underlying tasks do not carry.

Where the literature currently sits

There is broad agreement that financial constraint and cognitive performance are related, and continuing disagreement about magnitude, about which kinds of constraint produce the effect, and about whether short-run liquidity fluctuations reproduce what the harvest study observed. Later work has moved toward larger samples and pre-registered designs, which is the appropriate response to a contested finding.

For anyone reading secondary coverage of this literature, the useful distinction is between the original claim, which was narrow and conditional, and its circulation, which has generally not been.

Sources

• Mani, A., Mullainathan, S., Shafir, E., & Zhao, J. (2013). Poverty Impedes Cognitive Function. Science, 341(6149).
• Carvalho, L., Meier, S., & Wang, S. (2016). Poverty and Economic Decision-Making: Evidence from Households on Payday. American Economic Review, 106(2).
• Wicherts, J. M., & Scholten, A. Z. (2013). Comment on “Poverty Impedes Cognitive Function.” Science, 342(6163).
• Mullainathan, S., & Shafir, E. (2013). Scarcity: Why Having Too Little Means So Much. Times Books.